pca + lda model

if(!require("EBImage")){
  install.packages("BiocManager")
  BiocManager::install("EBImage")
}
if(!require("R.matlab")){
  install.packages("R.matlab")
}
if(!require("readxl")){
  install.packages("readxl")
}

if(!require("dplyr")){
  install.packages("dplyr")
}
if(!require("readxl")){
  install.packages("readxl")
}

if(!require("ggplot2")){
  install.packages("ggplot2")
}

if(!require("caret")){
  install.packages("caret")
}

if(!require("glmnet")){
  install.packages("glmnet")
}

if(!require("WeightedROC")){
  install.packages("WeightedROC")
}

if(!require("MASS")){
  install.packages("MASS")
}

install.packages('ROSE')

trying URL 'https://cran.rstudio.com/bin/macosx/contrib/4.0/ROSE_0.0-3.tgz'
Content type 'application/x-gzip' length 111353 bytes (108 KB)
==================================================
downloaded 108 KB

The downloaded binary packages are in
    /var/folders/h6/g930s4b173zglxv0sl72b5zr0000gn/T//RtmpYsmB72/downloaded_packages

library(R.matlab)
library(readxl)
library(dplyr)
library(EBImage)
library(ggplot2)
library(caret)
library(glmnet)
library(WeightedROC)
library(MASS)
library(ROSE)

Step 0 set work directories

set.seed(2020)
setwd('/Users/zy/Documents/GitHub/Spring2021-Project3-group-5/doc')
# here replace it with your own path or manually set it in RStudio to where this rmd file is located. 
# use relative path for reproducibility

Provide directories for training images. Training images and Training fiducial points will be in different subfolders.

train_dir <- "../data/train_set/" # This will be modified for different data sets.
train_image_dir <- paste(train_dir, "images/", sep="")
train_pt_dir <- paste(train_dir,  "points/", sep="")
train_label_path <- paste(train_dir, "label.csv", sep="") 

Step 1: set up controls for evaluation experiments.

In this chunk, we have a set of controls for the evaluation experiments.

• (T/F) cross-validation on the training set
• (T/F) reweighting the samples for training set
• (number) K, the number of CV folds
• (T/F) process features for training set
• (T/F) run evaluation on an independent test set
• (T/F) process features for test set

run.cv <- TRUE # run cross-validation on the training set
sample.reweight <- TRUE # run sample reweighting in model training
K <- 5  # number of CV folds
run.feature.train <- TRUE # process features for training set
run.test <- TRUE # run evaluation on an independent test set
run.feature.test <- TRUE # process features for test set

Step 2: import data and train-test split

#train-test split
info <- read.csv(train_label_path)
n <- nrow(info)
n_train <- round(n*(4/5), 0)
train_idx <- sample(info$Index, n_train, replace = F)
test_idx <- setdiff(info$Index, train_idx)

If you choose to extract features from images, such as using Gabor filter, R memory will exhaust all images are read together. The solution is to repeat reading a smaller batch(e.g 100) and process them.

n_files <- length(list.files(train_image_dir))

image_list <- list()
for(i in 1:100){
   image_list[[i]] <- readImage(paste0(train_image_dir, sprintf("%04d", i), ".jpg"))
}

Fiducial points are stored in matlab format. In this step, we read them and store them in a list.

#function to read fiducial points
#input: index
#output: matrix of fiducial points corresponding to the index
readMat.matrix <- function(index){
     return(round(readMat(paste0(train_pt_dir, sprintf("%04d", index), ".mat"))[[1]],0))
}

#load fiducial points
fiducial_pt_list <- lapply(1:n_files, readMat.matrix)
save(fiducial_pt_list, file="../output/fiducial_pt_list.RData")

Step 3: construct features and responses

• The follow plots show how pairwise distance between fiducial points can work as feature for facial emotion recognition.

  • In the first column, 78 fiducials points of each emotion are marked in order.
  • In the second column distributions of vertical distance between right pupil(1) and right brow peak(21) are shown in histograms. For example, the distance of an angry face tends to be shorter than that of a surprised face.
  • The third column is the distributions of vertical distances between right mouth corner(50) and the midpoint of the upper lip(52). For example, the distance of an happy face tends to be shorter than that of a sad face.

Figure1

feature.R should be the wrapper for all your feature engineering functions and options. The function feature( ) should have options that correspond to different scenarios for your project and produces an R object that contains features and responses that are required by all the models you are going to evaluate later.

• feature.R
• Input: list of images or fiducial point
• Output: an RData file that contains extracted features and corresponding responses

source("../lib/feature.R")
tm_feature_train <- NA
if(run.feature.train){
  tm_feature_train <- system.time(dat_train <- feature(fiducial_pt_list, train_idx))
  save(dat_train, file="../output/feature_train.RData")
}else{
  load(file="../output/feature_train.RData")
}

tm_feature_test <- NA
if(run.feature.test){
  tm_feature_test <- system.time(dat_test <- feature(fiducial_pt_list, test_idx))
  save(dat_test, file="../output/feature_test.RData")
}else{
  load(file="../output/feature_test.RData")
}

#rebalance the training data
balanced_train_data <- ROSE(label~.,data = dat_train)$data
save(balanced_train_data, file="../output/feature_balanced_train.RData")

Step 4: Train a classification model with training features and responses

Since there are over 6000 features, we implement the PCA method to reduce dimension according to the covariance matrix. We only retain PCs with large variance.

#separate the features from label
dat_train_new <- balanced_train_data[,-dim(balanced_train_data)[2]]
dat_test_new <- dat_test[,-dim(dat_test)[2]]
#create a vector contain target number of PCs
num.pca <- c(10,50,500,1000)

train_pca <- function(num.pca){
  for(i in 1:length(num.pca)){
    #start time for training the model
    train.model.start = proc.time()
    #run PCA
    pca <- prcomp(dat_train_new)
    #store for each potential PC
    train_pca <- data.frame(pca$x[,1:num.pca[i]], label = balanced_train_data[dim(balanced_train_data)[2]])
    pred_pca=predict(pca,dat_test_new)
    test_pca=data.frame(pred_pca[,1:num.pca[i]], label = dat_test[dim(dat_test)[2]])
    #fitting the lda model 
    lda_pca <- lda(label ~ ., data = train_pca) 
    #stop time for training the model
    train.model.end = proc.time()
    #start time for testing the model
    test.model.start = proc.time()
    #predict lda model
    lda_pred_pca = predict(lda_pca,test_pca[-dim(test_pca)[2]])
    #end time for testing the model
    test.model.end = proc.time()
    #test accuracy
    test_accuracy=confusionMatrix(lda_pred_pca$class, test_pca$label)$overall[1]
    print(list(l1=train.model.end - train.model.start,
               l2=test.model.end - test.model.start,
               l3=test_accuracy))
    }
  }
train_pca(num.pca)

$l1
   user  system elapsed 
165.321   1.728 168.072 

$l2
   user  system elapsed 
  0.002   0.000   0.002 

$l3
 Accuracy 
0.6183333 

$l1
   user  system elapsed 
156.726   1.163 158.476 

$l2
   user  system elapsed 
  0.008   0.007   0.018 

$l3
 Accuracy 
0.6816667 

$l1
   user  system elapsed 
159.786   1.354 161.673 

$l2
   user  system elapsed 
  0.027   0.004   0.031 

$l3
 Accuracy 
0.7033333 

$l1
   user  system elapsed 
180.074   1.533 182.535 

$l2
   user  system elapsed 
  0.064   0.006   0.071 

$l3
 Accuracy 
0.7116667 

By comparing the training time, test time and accuracy, we use model with 50 PCs.

pca_10 <- prcomp(dat_train_new)
train_pca_10 <- data.frame(pca_10$x[,1:50], label = balanced_train_data[dim(balanced_train_data)[2]])
pred_pca_10 <- predict(pca_10,dat_test_new)
test_pca_10 <- data.frame(pred_pca_10[,1:50], label = dat_test[dim(dat_test)[2]])

save(train_pca_10, file="../output/feature_pca_train.RData")
save(test_pca_10, file="../output/feature_pca_test.RData")

tm_train = NA
tm_train <- system.time(lda_pca_10 <- lda(label ~ ., data = train_pca_10,cv = TRUE))

Calculate the training and testing accuracy of LDA model

pred_train_lda <- predict(lda_pca_10, train_pca_10[-dim(train_pca_10)[2]])
accu_train_lda <- mean(pred_train_lda$class == train_pca_10$label)
cat("The trainig accuracy of model: LDA", "is", accu_train_lda*100, "%.\n")

The trainig accuracy of model: LDA is 71.70833 %.

tm_test = NA
if(run.test){
  tm_test <- system.time(pred_test_lda <- predict(lda_pca_10, test_pca_10))
}

save(pred_test_lda, file="../output/fit_train.RData")

accu_test_lda <- mean(pred_test_lda$class == test_pca_10$label)
cat("The accuracy of model: LDA", "is", accu_test_lda*100, "%.\n")

The accuracy of model: LDA is 68.16667 %.

Summarize Running Time

Prediction performance matters, so does the running times for constructing features and for training the model, especially when the computation resource is limited.

cat("Time for constructing training features=", tm_feature_train[1], "s \n")

Time for constructing training features= 1.059 s 

cat("Time for constructing testing features=", tm_feature_test[1], "s \n")

Time for constructing testing features= 0.198 s 

cat("Time for training model=", tm_train[1], "s \n") 

Time for training model= 0.061 s 

cat("Time for testing model=", tm_test[1], "s \n")

Time for testing model= 0.004 s 

###Reference - Du, S., Tao, Y., & Martinez, A. M. (2014). Compound facial expressions of emotion. Proceedings of the National Academy of Sciences, 111(15), E1454-E1462.